Relaxation in an L-infinity-optimization problem